INDEX

Explanations

used as input for training

New Auto-Interp

Top Features by Cosine Similarity

Configuration

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Negative Logits

âĦĸâĦĸ

-0.10

utow

-0.10

strand

-0.10

 Tier

-0.09

pires

-0.09

Typ

-0.09

ãĤ¿ãĥ«

-0.09

 ButterKnife

-0.08

comed

-0.08

 Admir

-0.08

POSITIVE LOGITS

 input

0.19

 feeding

0.18

 model

0.17

 training

0.17

 machine

0.16

feeding

0.15

fed

0.14

 Input

0.14

 feed

0.14

input

0.13

Activations Density 0.053%

No Known Activations